Press "Enter" to skip to content

How can AI help our NHS and should we be concerned? | Rebecca Pope | TEDxFolkestone


[Music]
your health is in your hands now perhaps
you’re thinking wait a minute I’ve come
along to a TED talk where I meant to
hear inspirational and pushing the
boundaries idea but I hear this every
time I go to the GP how is that a novel
idea but I don’t mean this in the
metaphorical sense but rather literally
there are more mobile devices than there
are people on the planet and every time
you pick up your phone your laptop your
tablet you’re interacting with
artificial intelligence perhaps because
you’re searching for something on Google
you’re uploading a photograph on
Facebook or asking Siri the meaning of
life you’re already interacting with
artificial intelligence you just may not
know it now before I talk about what
artificial intelligence is and how these
systems crucially learn I want to talk
about what artificial intelligence is
not please expunge this idea from your
mind
[Laughter]
this is the media’s portrayal of
artificial intelligence and it comes
about because we anthropomorphic Isis
technology we try and humanize this
digital technology artificial
intelligence is not the same as human
intelligence these systems don’t enjoy
this beautiful inner mental world that
we have of consciousness so what is AI
well you know because you use it every
day it’s a tool and since our inception
I love this slide to my favorite person
since our inception we have been master
tool makers culminating in the 1940s
with the advent of the computer now for
many years human programmers people like
me you wouldn’t get invited to many
dinners would sit behind computers and
history and handcraft code into the
machine instructions to tell it what to
do and we did pretty well we we got
expert system systems that could outwit
humans at particular tasks and it
actually 20 years ago today would by
sheer coincidence that IBM’s deep blue
beat the then world champion Garry
Kasparov at his own game
at chess but then in the 1980s there was
a revolution in computer science and it
was an introduction of pac-man though
that was awesome but rather the
introduction of computer programs that
could learn so how how can computers
learn well actually the inspiration
comes from the most magnificent machine
we will ever own our own brains so if we
think about how infants learn like this
QT behind me not my son but you know
I’ve got the image of a good website so
it’s fine um if you think about how
infants learn they learn from experience
they learn from trial and error so every
time this gorgeous little one goes to
play with this guy
new connections will form in the brain
and over time those connections will
become more efficient they’ll be able to
talk to other parts of the brain better
he’ll acquire the motor and cognitive
skills he needs to be able to orient the
blocks into the right spaces that’s how
he’s learned and this is the idea that
we do in artificial intelligence the
computers have experiences well yes they
do they have data so what we do is we
allow AI systems to learn from data
without being pre-programmed now this
all might seem a little bit abstract and
a bit sci-fi which I don’t want to do so
I wanted to show you what researchers at
Google had created an artificial
intelligence system to play a 1980s game
called breakout now the only one piece
of code in this AI system is the goal
and that goal was to maximize the score
on the game how the computer would get
the highest score what it would need to
do it didn’t know it had been
pre-programmed into the system now a
hundred games you can see that it’s okay
it’s had a bit experience it kind of
knows what it’s doing but then it gets
more experience more data and 300 games
you can see that it’s kind of a human
capabilities it’s knowing to move the
bat left and right to keep the ball in
play and that will knock the blocks and
it’s cool go up and then quite
unexpectedly they thought well let’s
just leave this overnight because
computer programmers do that because
we’re lazy and let’s see what let’s see
what happens and it learnt for itself
this optimal solution to use the ball
how cool is that right if only I had a
brain like that to drill up the
left-hand side and to get the highest
score in as fewer possible moves as it
could and it’s this technology this
ability for computer programs to learn
from data and we’re in a data fueled
world where are we find ourselves in
this
officially intelligent world that we now
live in and this technology has
revolutionized a number of sectors
banking marketing insurance and the list
goes on but there’s one sector and in my
opinion the most important sector where
this technology is not being utilized to
its full potential and that is
healthcare now we know that our
wonderful NHS collects an awful lot of
information about us and they they
either reside in unhackable paper that
wasn’t me either I’m happy we’ll paid
for files like this or in disparate
clinical systems that don’t necessarily
talk to each other so the compelling
question is how do we develop and deploy
artificial intelligence systems to learn
from our data to ultimately help our
very are struggling NHS system our
healthcare system and I want to be
really clear here when I say deploy I do
not mean replace our wonderful doctors
nurses therapists but rather give them
tools artificial intelligence tools to
make their working lives more proficient
and ultimately let them care for
patients because that’s why people
joined the NHS now 1 million of us use
our NHS every 36 hours and for the vast
majority of us we receive world-class
care but sadly for one in ten of us
during a hospital mission we will
undergo a medical error or a medical
harm and research indicates that one in
20 deaths in hospital is preventable so
how can we start to use the digital
technologies that you probably hopefully
I’m playing on now but have in your
pockets to reduce these figures and help
our healthcare system one way is digital
or health we all probably have mobiles
on us right now and those Mobile’s may
be very useful for example in ordering
on Sainsbury’s restroom and slightly
higher tax bracket weight trays and um
and also in our dating so these apps are
everywhere but we don’t use them in our
healthcare why not they’re very simple
to use and it’s this idea only last
month that the NHS adopted they
developed this website which is a
digital tool for people with a multitude
of long-term conditions you can see here
for example diabetes dementia cancer
living in the community to use these
apps to help them manage their condition
and I want to just take a deep dive into
one of these apps which was developed by
researchers and clinicians at Oxford
University
now patients were given a tablet not a
medical one one like this eight that’s a
proper pun they were given this in their
own homes and they were given a probe to
pop on their finger and every day that
that probe would read data about that
patient vital signs like heart rate the
amount of oxygen in the blood and over
time that system the artificial
intelligence system under the hood of
this app would learn about this patient
because even though you me may have a
diagnosis we are all different our
diagnosis is one thing but we are might
be very different in the way we react in
the way we integrate with our with our
condition so what do these apps allow
for well I argue that they allow for a
more proactive rather than reactive
healthcare system so let’s think about
if we had a fictional patient with for
example COPD which is chronic
obstructive pulmonary disease affects 3
million people in the UK long term
condition and they became acutely
breathless what options they have at the
moment well they could wring NHS 1 1 1
to get
advice or they could ring 999
understandably both of those options
were probably culminate in an ambulance
being dispatched then being rushed there
Lake Laney hospital being admitted
patients don’t want to be in hospital
they want to be empowered in their own
health and how they look after
themselves but let’s say that these apps
were being used
well that would allow for a proactive
health care system rather than reactive
one because these apps for example this
particular one is able to predict ahead
of time way before the patient starts to
notice breathlessness feelings that
they’re becoming unwell and they’re on
this trajectory of health deterioration
and what this app can then do is alert
the clinical team so the patient’s
doctor or spiritually nurse I think you
need to check on this patient and so we
then start to get personalized treatment
plans because as I said no two patients
with copd are the same now this
particular app has been running for 12
months and showing really promising
results patients using this app in
Oxford there’s been a reduction of 17%
of hospital admissions and 28% of GP
visits have been reduced but that’s the
stat that’s that’s more important what’s
important is that patients were
empowered in their own health they
literally had their health in their
hands and they were a collaborator with
their clinical team through this digital
technology to look after themselves
now the second example is one very close
to my heart I’ve for 25 years now I
don’t look at type 1 diabetes and a
worry for patients like me is is I sight
loss so patients like me are 25 times
more likely to suffer site loss and
eventual blindness if caught early
enough ninety-eight percent of these
cases could be prevented so what’s the
current state of play well an
ophthalmologist an eye doctor will take
a picture like this with the back of my
eye this isn’t the back of my eye
thankfully but and we’ll start to troll
through these scans
looking for features or characteristics
that indicate that I have diabetic eye
disease now this is a very very
time-intensive thing for a doctor to do
and we kind of have a perfect storm at
the moment because we have very few
ophthalmologists it takes a long time to
become a consultant eye doctor but at
the same time the number of people being
diagnosed with diabetes in the UK is
rapidly rising so a leading Hospital in
London Moorfields Eye Hospital teamed up
with google deepmind and said can we
train an artificially intelligent system
to read these scans now this hasn’t been
integrated into the NHS it’s still in a
research phase but if it were successful
what are the kind of benefits it could
give us well it would be able to triage
patients according to their need so at
the moment doctors see scans and they
don’t know what they’re going to see so
a patient might be 99th in the row that
is probably going to go blind in the
next month versus someone he’s looking
at or she’s looking at at that
particular day and it would take him
along or her a long time to get to that
patient that is really in need of their
treatment but because as we saw earlier
we don’t turn off these systems they can
run all night they don’t go on annual
leave they can start to triage these
patients according to need now again as
I said earlier we’re not replacing
doctors but rather streamlining their
workloads so that the right patients are
seen at the right time so with all this
wonderful technology that we have
available and we’re trying to infuse
into the NHS and into our healthcare
system there are big players moving into
this space Google Microsoft as we know
IBM and this ecosystem is growing
rapidly so should we be concerned not
it’s my opinion that we need to partner
with third parties for the NHS to
continue to offer world-class care
particularly under the operational
financial pressures it currently finds
itself in but there are ethical and
regulatory concerns that we need to
discuss in public forums like this for
example only yesterday as I’m sure we
all know there was a cyber attack on NHS
systems if we’re sharing data between
the NHS and other other operational
systems for example let’s say in KPMG
your IBM or wherever how do we make sure
that that data isn’t hackable if we give
our data to private companies who owns
that data and what can they do with it
how do we ensure that our healthcare
data unlike our facebook data doesn’t
become a monetized commodity for example
I don’t want to go onto the internet and
be you know looking at the latest cat
video and and I mean do my work and and
this is recorded right I’m gonna start
to see advertisements for diabetic
related products I find it really spooky
in it and very intrusive and we heard
earlier that these artificial
intelligence systems can allow doctors
beep and other members of the clinical
team to be more effective in their
diagnosis but what happens because no
system is perfect even a digital one
when the system gets it wrong when it
misses that cancerous growth from a
legal perspective who is responsible me
the developer the doctor or the system
itself which as I said earlier isn’t
human
where is it gray gray area and we need
to discuss this as a society so it would
be remiss of me as a data scientist not
to predict something and how do if we if
we’re able to tackle these very
important and I think game-changing
questions where do I see the this new
algorithmic healthcare landscape passing
what will it look like well I think if
we do this
Polly and effectively and in
collaboration with the NHS then
paradoxically our NHS could actually
become more human more personal because
doctors nurses and therapists time will
be freed up and no algorithm or
artificial intelligence system can
comfort you or be there for you
when your worst fears are confirmed
empathy is and I believe will always
remain a uniquely human trait thank you
for your time
[Applause]
[Applause]
Please follow and like us: